def train(self, images, labels):
"""
Train eigenfaces
"""
print "Train..."
#copy labels
self._labels = labels
#transform the numpe vector to shogun structure
features = RealFeatures(images)
#PCA
self.pca = PCA()
#set dimension
self.pca.set_target_dim(self._num_components)
#compute PCA
self.pca.init(features)
for sampleIdx in range(features.get_num_vectors()):
v = features.get_feature_vector(sampleIdx)
p = self.pca.apply_to_feature_vector(v)
self._projections.insert(sampleIdx, p)
print "Train ok!"
def preprocessor_pca_modular (data): from modshogun import RealFeatures from modshogun import PCA features = RealFeatures(data) preprocessor = PCA() preprocessor.init(features) preprocessor.apply_to_feature_matrix(features) return features
def RunPCAShogun(q):
totalTimer = Timer()
# Load input dataset.
Log.Info("Loading dataset", self.verbose)
try:
feat = RealFeatures(self.data.T)
with totalTimer:
# Get the options for running PCA.
if "new_dimensionality" in options:
k = int(options.pop("new_dimensionality"))
if (k > self.data.shape[1]):
Log.Fatal("New dimensionality (" + str(k) +
") cannot be greater than" +
"existing dimensionality (" +
str(self.data.shape[1]) + ")!")
q.put(-1)
return -1
else:
k = self.data.shape[1]
if "whiten" in options:
s = True
options.pop("whiten")
else:
s = False
if len(options) > 0:
Log.Fatal("Unknown parameters: " + str(options))
raise Exception("unknown parameters")
# Perform PCA.
prep = ShogunPCA(s)
prep.set_target_dim(k)
prep.init(feat)
prep.apply_to_feature_matrix(feat)
except Exception as e:
q.put(-1)
return -1
time = totalTimer.ElapsedTime()
q.put(time)
return time
def train(self, images, labels):
"""
Train eigenfaces
"""
print "Train...",
#copy labels
self._labels = labels;
#transform the numpe vector to shogun structure
features = RealFeatures(images)
#PCA
self.pca = PCA()
#set dimension
self.pca.set_target_dim(self._num_components);
#compute PCA
self.pca.init(features)
for sampleIdx in range(features.get_num_vectors()):
v = features.get_feature_vector(sampleIdx);
p = self.pca.apply_to_feature_vector(v);
self._projections.insert(sampleIdx, p);
print "ok!"
def RunPCAShogun(q):
totalTimer = Timer()
# Load input dataset.
Log.Info("Loading dataset", self.verbose)
try:
feat = RealFeatures(self.data.T)
with totalTimer:
# Find out what dimension we want.
match = re.search('-d (\d+)', options)
if not match:
k = self.data.shape[1]
else:
k = int(match.group(1))
if (k > self.data.shape[1]):
Log.Fatal("New dimensionality (" + str(k) +
") cannot be greater than" +
"existing dimensionality (" +
str(self.data.shape[1]) + ")!")
q.put(-1)
return -1
# Get the options for running PCA.
s = True if options.find("-s") > -1 else False
# Perform PCA.
prep = ShogunPCA(s)
prep.set_target_dim(k)
prep.init(feat)
prep.apply_to_feature_matrix(feat)
except Exception as e:
q.put(-1)
return -1
time = totalTimer.ElapsedTime()
q.put(time)
return time
def RunPCAShogun():
totalTimer = Timer()
# Load input dataset.
Log.Info("Loading dataset", self.verbose)
try:
feat = RealFeatures(self.data.T)
with totalTimer:
# Get the options for running PCA.
if "new_dimensionality" in options:
k = int(options.pop("new_dimensionality"))
if (k > self.data.shape[1]):
Log.Fatal("New dimensionality (" + str(k) + ") cannot be greater than"
+ "existing dimensionality (" + str(self.data.shape[1]) + ")!")
return -1
else:
k = self.data.shape[1]
if "whiten" in options:
s = True
options.pop("whiten")
else:
s = False
if len(options) > 0:
Log.Fatal("Unknown parameters: " + str(options))
raise Exception("unknown parameters")
# Perform PCA.
prep = ShogunPCA(s)
prep.set_target_dim(k)
prep.init(feat)
prep.apply_to_feature_matrix(feat)
except Exception as e:
return -1
return totalTimer.ElapsedTime()
def RunPCAShogun(q):
totalTimer = Timer()
# Load input dataset.
Log.Info("Loading dataset", self.verbose)
try:
feat = RealFeatures(self.data.T)
with totalTimer:
# Find out what dimension we want.
match = re.search('-d (\d+)', options)
if not match:
k = self.data.shape[1]
else:
k = int(match.group(1))
if (k > self.data.shape[1]):
Log.Fatal("New dimensionality (" + str(k) + ") cannot be greater than"
+ "existing dimensionality (" + str(self.data.shape[1]) + ")!")
q.put(-1)
return -1
# Get the options for running PCA.
s = True if options.find("-s") > -1 else False
# Perform PCA.
prep = ShogunPCA(s)
prep.set_target_dim(k)
prep.init(feat)
prep.apply_to_feature_matrix(feat)
except Exception as e:
q.put(-1)
return -1
time = totalTimer.ElapsedTime()
q.put(time)
return time
class EigenFaces():
def __init__(self, num_components):
"""
Constructor
"""
self._num_components = num_components
self._projections = []
def train(self, images, labels):
"""
Train eigenfaces
"""
print "Train..."
#copy labels
self._labels = labels
#transform the numpe vector to shogun structure
features = RealFeatures(images)
#PCA
self.pca = PCA()
#set dimension
self.pca.set_target_dim(self._num_components)
#compute PCA
self.pca.init(features)
for sampleIdx in range(features.get_num_vectors()):
v = features.get_feature_vector(sampleIdx)
p = self.pca.apply_to_feature_vector(v)
self._projections.insert(sampleIdx, p)
print "Train ok!"
def predict(self, image):
"""
Predict the face
"""
#image as row
imageAsRow = np.asarray(
image.reshape(image.shape[0] * image.shape[1], 1), np.float64)
#project inthe subspace
p = self.pca.apply_to_feature_vector(
RealFeatures(imageAsRow).get_feature_vector(0))
#min value to find the face
minDist = 1e100
#class
minClass = -1
#search which face is the best match
for sampleIdx in range(len(self._projections)):
test = RealFeatures(np.asmatrix(p, np.float64).T)
projection = RealFeatures(
np.asmatrix(self._projections[sampleIdx], np.float64).T)
dist = EuclideanDistance(test, projection).distance(0, 0)
if (dist < minDist):
minDist = dist
minClass = self._labels[sampleIdx]
return minClass
def getMean(self):
"""
Return the mean vector
"""
return self.pca.get_mean()
def getEigenValues(self):
"""
Return the eigenvalues vector
"""
return self.pca.get_eigenvalues()
#Reconstruction with diferents values of eigenvectos
#Read the last image of the file to test Eigenfaces
image = cv2.resize(cv2.imread(list_filenames[0], cv2.IMREAD_GRAYSCALE),
(IMAGE_HEIGHT, IMAGE_WIDHT))
#image as row
imageAsRow = np.asarray(image.reshape(image.shape[0] * image.shape[1], 1),
np.float64)
#Reconstruct 10 eigen vectors to 300, step 15
for i in range(10, 300, 50):
print "Reconstruct with " + str(i) + " eigenvectors"
pca = PCA()
#set dimension
pca.set_target_dim(i)
#compute PCA
pca.init(RealFeatures(images))
pca.apply_to_feature_vector(
RealFeatures(imageAsRow).get_feature_vector(0))
#reconstruct
projection = pca.apply_to_feature_vector(
RealFeatures(imageAsRow).get_feature_vector(0))
reconstruction = np.asmatrix( np.asarray(projection, np.float64))* \
np.asmatrix( pca.get_transformation_matrix()).T
reconstruction = reconstruction + pca.get_mean()
class EigenFaces():
def __init__(self, num_components):
"""
Constructor
"""
self._num_components = num_components;
self._projections = []
def train(self, images, labels):
"""
Train eigenfaces
"""
print "Train...",
#copy labels
self._labels = labels;
#transform the numpe vector to shogun structure
features = RealFeatures(images)
#PCA
self.pca = PCA()
#set dimension
self.pca.set_target_dim(self._num_components);
#compute PCA
self.pca.init(features)
for sampleIdx in range(features.get_num_vectors()):
v = features.get_feature_vector(sampleIdx);
p = self.pca.apply_to_feature_vector(v);
self._projections.insert(sampleIdx, p);
print "ok!"
def predict(self, image):
"""
Predict the face
"""
#image as row
imageAsRow = np.asarray(image.reshape(image.shape[0]*image.shape[1],1),
np.float64);
#project inthe subspace
p = self.pca.apply_to_feature_vector(RealFeatures(imageAsRow).get_feature_vector(0));
#min value to find the face
minDist =1e100;
#class
minClass = -1;
#search which face is the best match
for sampleIdx in range(len(self._projections)):
test = RealFeatures(np.asmatrix(p,np.float64).T)
projection = RealFeatures(np.asmatrix(self._projections[sampleIdx],
np.float64).T)
dist = EuclideanDistance( test, projection).distance(0,0)
if(dist < minDist ):
minDist = dist;
minClass = self._labels[sampleIdx];
return minClass
def getMean(self):
"""
Return the mean vector
"""
return self.pca.get_mean()
def getEigenValues(self):
"""
Return the eigenvalues vector
"""
return self.pca.get_eigenvalues();
image = cv2.resize(cv2.imread(list_filenames[0], cv2.IMREAD_GRAYSCALE),
(IMAGE_HEIGHT, IMAGE_WIDHT), interpolation=cv2.INTER_LINEAR);
#image as row
imageAsRow = np.asarray(image.reshape(image.shape[0]*image.shape[1],1),
np.float64);
reconstructions = range(10, 250, 50)
reconstructions_images = np.empty( (IMAGE_HEIGHT, IMAGE_WIDHT*len(reconstructions) ), np.uint8)
#Reconstruct 10 eigen vectors to 300, step 15
for i in reconstructions:
print "Reconstruct with " + str(i) + " eigenvectors"
pca = PCA()
#set dimension
pca.set_target_dim(i);
#compute PCA
pca.init(RealFeatures(images))
pca.apply_to_feature_vector(RealFeatures(imageAsRow)
.get_feature_vector(0));
#reconstruct
projection = pca.apply_to_feature_vector(RealFeatures(imageAsRow)
.get_feature_vector(0));
reconstruction = np.asmatrix( np.asarray(projection, np.float64))* \
np.asmatrix( pca.get_transformation_matrix()).T
reconstruction = reconstruction + pca.get_mean()
